Once an utterance is divided syntagmatically into a sequence of recurrent partials, primary as well as secondary, these elements must be grouped paradigmatically into abstract morphemes. The three basic criteria that guide this classification are set out in Harris (1942):

A morpheme unit is thus a group of one or more alternants which have the same meaning and complementary distribution.

In units consisting of more than one alternant, the total distribution of all the alternants (i.e. the combined range of environments in which each of them occurs) must equal the range of environments in which some unit with but a single alternant occurs. (Harris

1942: 171)

This combination of semantic and distributional criteria were largely carried forward in subsequent versions of Post-Bloomfieldian morphemic analysis. The principal revision introduced in Hockett (1947) replaces the requirement that morphs must occur in “complementary” distribution by the weaker condition that their distribution must merely be “non-contrastive”:

Two or more morphs are grouped into a single morpheme if they fit the following grouping requirements: (a) they have the same meaning; (b) they are in non-contrastive distribution; and (c) the range of the resultant morpheme is not unique. (Hockett 1947:322f.)

The first criterion imposed by Harris (1942) and Hockett (1947) characterizes morphemes as semantically (rather than formally) coherent classes of elements. Hence although meaning is not part of a grammatical description per se in the IA model, meaning (or at least a contrast between sameness and difference of meaning) is fundamental to morphemic analysis. The second criterion in Harris (1942) requires that two realizations of a common morpheme must be in complementary distribution. The revision in Hockett (1947) reflects his view that this condition is too strong in cases where alternatives occur with no obvious difference in meaning or function. English noun plurals provide a simple illustration. Hockett notes that the exponent [s] and the exponent that consists of final segment voicing together with [z] occur in non-complementary distribution in pairs such as hoofs^hooves. Pairs such as oxen^oxes show a similarly non-complementary distribution for [on] and [sz] To account for these cases, Hockett weakens Harris’s second criterion to the requirement that morphs need only occur in non-contrastive distribution.

The third criterion is the most subtle. Its most direct effect is to prevent synonymous forms from being coerced into morphemes. Harris (1942) illustrates this effect by contrasting the morphemic status of the forms of English be with the synonyms twenty and (somewhat archaic) score:

Thus the combined environments of am, are, be are included in the environments in which walk occurs: I am, they are, to be, as compared with I walk, they walk, to walk. The case is different with twenty and score, even though they have the same meaning and never occur in the same environment. For there is no morpheme unit in English which consists of only one alternant and which occurs in the combined distribution of twenty and score. Therefore, we consider the alternants am, are, be as being members of a single morpheme unit; but of the alternants twenty and score, each constitutes a morpheme unit by itself. (Harris 1942:172)

The third criterion has a similar, though less obvious, effect on the grouping of inflectional exponents. On any morphemic analysis, the English plural morpheme {s} will contain the phonologically conditioned allomorphs [z], [sz] and [s]. However, the status of the [sn] in oxen or the ‘0’ associated with sheep is less straightforward. These allomorphs are clearly morphologically conditioned, given that phonologically similar nouns take the regular plural by default, as in the case of box^ boxes and heap^heaps. The treatment of these elements ultimately distinguishes two different conceptions of the morpheme. If [sn] and ‘0’ are not assigned to the morpheme {s}, then its realizations will retain a similarity in form.

Morphological and phonological allomorphy

Figure 2.1 Morphological and phonological allomorphy

If formally heterogeneous allomorphs such as [эп] and ‘0’ are included, then {s} will merely enumerate all of the strategies for expressing plural number in English. All of the candidate allomorphs of {s} satisfy the first two conditions specified by Harris (1942) and Hockett (1947): they all have the same meaning and occur in a non-contrastive distribution.

It is the third criterion that is decisive here. If {s} isto have a non-unique range, then some morpheme must have the same distribution as the plural allomorphs do collectively. Harris (1942:111) suggests that “the range of environments” of {s} “equals that of zero ‘singular, the suffix -ful and other single-alternant morpheme units”. Yet neither of Harris’s examples is persuasive. As Hockett (1947:230) notes, “[t]he zero element with meaning ‘noun singular’... has a very dubious status, having no alternant of other than zero shape”. Likewise even if one accepts that -ful or some other derivational affix combines as productively with noun stems as plural exponents do, it is unclear why the distribution of a derivational ending should be relevant to the identification of an inflectional morpheme. Later refinements of morphemic analysis do not improve on Harris’s third criterion. For example, the far more elaborate conditions in Nida (1948:421) reintroduce a version of this criterion as a restriction on the grouping of “[f]orms which possess a common semantic distinctiveness, but which differ in their phonemic form”.

The issues that arise in determining whether [un] is an allomorph of {s} illustrate the kinds of problems created by morphemic analysis, even when applied to extremely simple patterns. Moreover, the genuine ambiguity present in the notions ‘morph’ and ‘allomorphy’ remains largely unresolved in later accounts. For the sake of clarity, it is therefore useful to reserve the term ‘morph’ for morphologically conditioned allomorphs, and apply Harris’ term ‘(morpheme) alternant’ to phonologically conditioned variants. This refinement of the Descriptivist terms is illustrated by the description of the English singular hoof, and the plurals hoofs and hooves in Figure 2.1. The lexical morpheme {hoof} has two lexically-conditioned morphs. The ‘default’ stem allomorph /huf/ realizes the singular hoof and underlies the regular plural formation hoofs. The voiced stem allomorph /huv/ occurs only in the alternative plural hooves. The grammatical plural morpheme {s} is realized in both plural forms by the allomorph /z/. This allomorph is in turn realized by the alternants [z] and [s] in the surface word forms hoofs and hooves.

< Prev   CONTENTS   Source   Next >